With the current crop of consoles from Microsoft and Sony nearing the end of their natural product cycles, tech rumours are abound as to what hardware and systems the next generation will have. Amongst a whole raft of claims as to what the PlayStation 6 will be like are a list of specifications for Sony’s return to the handheld market, with a beefy custom AMD chip at the heart of it all.
Now, before I go any further, let me get one thing out of the way first, and it’s the source of these claims: Moore’s Law is Dead. The tech YouTube channel’s modus operandi is all about rumours, leaks, speculation, and at times, pretty wild predictions. But hey, even if you spray about in a raging gale, something will eventually land on target.
On to the nitty-gritty, then. MLID claims Sony is planning a handheld PlayStation for its PS6 portfolio. Not a major shock announcement, as the company has done this before. Something else that won’t raise any eyebrows is that it’s apparently going to be powered by a custom AMD chip, codenamed Canis.
The specs for it, though, are a tad more of a surprise. Manufactured on TSMC’s N3 process node and coming in at 135 square millimetres in size, the CPU size of Canis is alleged to have four Zen 6c cores and two Zen 6 Low Power cores. That’s a little bit like AMD’s Ryzen AI 340, which sports two Zen 5 and four Zen 5c cores. However, unlike that laptop APU, MLID is suggesting that games will run on the 6c pipelines, with the handheld’s operating system being handled by the two LP cores.
There are no architectural differences between AMD’s normal Zen and Zen-c cores (at least, not in Zen 5) other than what clock speeds they can reach, but given that it’s also being claimed that the ‘PS6 handheld’ will be backwards compatible with PS5/PS4 games, I’m not sure how four, low-clocked cores are supposed to handle software designed for up to eight cores.
And that’s before one begins to question why Sony would choose to go with an architecture that AMD hasn’t released yet, when it’s historically chosen an older design that’s well-tested, proven, and predictable. Oh, and cheap. Very cheap.
Moving onto the GPU section of the APU, Canis is supposed to have 16 RDNA 5 compute units. To put this into perspective, the Steam Deck has eight RDNA 2 CUs, and the Asus ROG Ally X has 12 RDNA 3 CUs, so the compute unit count isn’t beyond the realms of possibility.
However, just as with the CPU section, I’m not overly convinced that Sony would go for what would be a cutting-edge GPU architecture for the release. Even the expensive PS5 Pro is still using what’s fundamentally an RDNA 2-powered GPU, albeit with some hefty modifications.
Where things get a bit silly are the claimed clock speeds and performance for the handheld’s GPU: around 1.65 GHz in docked mode and up to 75% of a PS5’s native rendering power. Sony’s current console has a GPU with 36 CUs, with a top clock speed of 2.2 GHz, and requires a power budget of 180 W.
While RDNA 5 rumours have yet to settle down into any semblance of sensibility, no amount of architectural wizardry can really overcome a 56% deficit in CUs with a 25% short fall in clock speed to that kind of degree. Well, perhaps it can, if the rendering resolution is low enough or the actual graphics workload leans more towards favouring AMD’s current shader design than for RDNA 2.
Just as with many handheld gaming PCs, Sony’s effort will apparently use LPDDR5X-8533, but rather than using a 128-bit wide bus, Canis is purported to sport a 192-bit bus, resulting in the total amount of RAM reaching 48 GB. That’s not impossible, as handhelds really do benefit from having considerably more than 16 GB of RAM, as it’s shared across the CPU and GPU.
Having watched MLID go through the specs, I was unconvinced by the CPU description, on the fence by the GPU (but not at all by the performance claims), and reasonably okay with the RAM specs. However, it wouldn’t be a MLID video if there wasn’t at least one really bonkers prediction, and in this instance, it’s the price: between $399 and $499.
The Switch 2’s retail price is $450, and it is a far weaker collection of hardware, with the main SoC manufactured on an old, cheap process node. Top-end handheld gaming PCs that are more akin to the above claimed specs are typically double the cost. Heck, even the PlayStation Portal is $200 and there’s practically nothing inside that beyond a basic Qualcomm chip, a smattering of RAM, and a pokey 16 Wh battery.
Sony wouldn’t set the price that low for a platform that isn’t going to sell anywhere near as many units as a normal console. It can afford to get away with a tiny profit margin with the PS5 because it hauls the money back via the millions of games sold each year for the console. A PS6 handheld would have to be physically profitable, and given that the specs are all next-gen architecture, on an expensive process node, $500 would surely be nowhere near enough.
Anyway, you can make up your own mind about MLID’s claims about the handheld or the other PS6 bits and pieces. Better yet, you can play your own game of ‘Guess the next-gen console specs’ and make a video of it, because everyone’s predictions will be just as valid as each other until the hardware itself finally appears.
Best handheld PC 2025
